Installing PRIMA Memory

The AI Business OS installer installs PRIMA Memory for you. You do not install it on its own, you do not edit any configuration file by hand, and you do not need a GitHub account.

Before You Start

Note

Check these three things before you paste the install prompt.

• Your order includes PRIMA Memory, since the installer adds only the components you own.
• Claude Code runs on your computer. The full list of what you need is on the Requirements page.
• Node.js 22.5 or later is on your computer, because PRIMA Memory uses the SQLite module built into those versions of Node.js.

Install PRIMA Memory

1. Open the email you received when you bought AI Business OS.
2. Open Claude Code in the workspace where you want session memory.
3. Copy the single install prompt from the email to your clipboard.
4. Paste the prompt into the Claude Code message box and send it.
5. Wait while Claude Code clones AI Business OS Core from git.genaiskills.io with the key in the prompt and runs the Core installer, which installs PRIMA Memory and your other components into the workspace.
6. Restart Claude Code when the installer finishes, so that it loads the new server.

Installing AI Business OS walks through the whole install, including the parts that are not specific to PRIMA Memory.

What the Installer Sets Up

Item	What it does
MCP server entry	Registers PRIMA Memory with Claude Code for this workspace, so the seven tools appear in every session you start there
PRIMA_EXPECTED_HOST set to claude-code	Fixes the server in Claude Code mode from the first launch, before the workspace has any saved sessions (see Configuration)
.prima-memory/ ignore rule	Keeps the index folder out of git, because the index holds a copy of your conversation text

PRIMA Memory creates the .prima-memory/ folder itself the first time it starts in the workspace.

Check That It Works

1. Start a new Claude Code session in the workspace where you installed it.
2. Ask Claude to run the PRIMA Memory get_index_status tool and show you the result.
3. Read the reply, which follows the layout of the example below with your own numbers.

Semantic index status:

Session files on disk: 42
Embedded: 0
Not yet embedded: 42

Exchanges by source:
  (index empty)
Total: 0 exchanges from 0 files

Rebuild job: none running

Coverage incomplete — 42 session file(s) are keyword-only. Run rebuild_index to embed them.

A reply of any kind means the server is running. A brand-new workspace shows zero session files, because Claude Code has not saved a conversation there yet. If Claude reports that it has no such tool, go to Troubleshooting.

Build the Search Index

Keyword search works straight away. Semantic search needs an index, and PRIMA Memory builds it in small steps before each search, for up to 20 seconds at a time. A workspace with a long history therefore takes several searches to reach full coverage.

To index your whole history in one go, ask Claude to run rebuild_index once after you install. The rebuild runs in the background while you carry on working, and Rebuilding the Index explains the process.

Note

The first semantic search or rebuild downloads the embedding model (about 23 MB) from Hugging Face. Your computer needs an internet connection for that one download. Privacy and Your Data describes exactly what the download sends.

Updating PRIMA Memory

Updates arrive through AI Business OS Core. When you run the AI Business OS update in the workspace, it refreshes every component you own, PRIMA Memory included. Restart Claude Code after the update so the new version takes effect. Updating Your Installation and the /update command page describe the update itself.

If you update from version 2.4 or earlier, PRIMA Memory deletes the old index file the first time it starts and creates a new one, because those versions stored vectors in a format the current version cannot read. The server writes a line saying so to its startup log. Your conversations are untouched, and the index fills again as you search or after a rebuild.

Removing PRIMA Memory

Uninstalling covers removing AI Business OS components. The data PRIMA Memory created lives in the .prima-memory/ folder in your workspace, and Privacy and Your Data explains how to delete it.

Developer Note: Running From a Clone

This route is for development of the server itself and is not the supported install for buyers.

git clone <repository> prima-memory
cd prima-memory
npm install
npm run build

Start the server from the workspace root, because the server takes the workspace from the folder it starts in. Use node <path-to-clone>/dist/index.js or the bundled start.sh launcher in an MCP server entry whose env block sets PRIMA_EXPECTED_HOST to claude-code, and add .prima-memory/ to the workspace .gitignore. Update a clone with git pull, npm install and npm run build.